systemd-boot expects being loaded from ESP and is quite unhappy in case
the loaded image device path is something else. When running on qemu
this can easily happen though. Case one is direct kernel boot, i.e.
loading via 'qemu -kernel systemd-bootx64.efi'. Case two is sd-boot
being added to the ovmf firmware image and being loaded from there.
This patch detects both cases and goes inspect all file systems known to
the firmware, trying to find the ESP. When present the
VMMBootOrderNNNN variables are used to inspect the file systems in the
given order.
Rather than passing seeds up to userspace via EFI variables, pass seeds
directly to the kernel's EFI stub loader, via LINUX_EFI_RANDOM_SEED_TABLE_GUID.
EFI variables can potentially leak and suffer from forward secrecy
issues, and processing these with userspace means that they are
initialized much too late in boot to be useful. In contrast,
LINUX_EFI_RANDOM_SEED_TABLE_GUID uses EFI configuration tables, and so
is hidden from userspace entirely, and is parsed extremely early on by
the kernel, so that every single call to get_random_bytes() by the
kernel is seeded.
In order to do this properly, we use a bit more robust hashing scheme,
and make sure that each input is properly memzeroed out after use. The
scheme is:
key = HASH(LABEL || sizeof(input1) || input1 || ... || sizeof(inputN) || inputN)
new_disk_seed = HASH(key || 0)
seed_for_linux = HASH(key || 1)
The various inputs are:
- LINUX_EFI_RANDOM_SEED_TABLE_GUID from prior bootloaders
- 256 bits of seed from EFI's RNG
- The (immutable) system token, from its EFI variable
- The prior on-disk seed
- The UEFI monotonic counter
- A timestamp
This also adjusts the secure boot semantics, so that the operation is
only aborted if it's not possible to get random bytes from EFI's RNG or
a prior boot stage. With the proper hashing scheme, this should make
boot seeds safe even on secure boot.
There is currently a bug in Linux's EFI stub in which if the EFI stub
manages to generate random bytes on its own using EFI's RNG, it will
ignore what the bootloader passes. That's annoying, but it means that
either way, via systemd-boot or via EFI stub's mechanism, the RNG *does*
get initialized in a good safe way. And this bug is now fixed in the
efi.git tree, and will hopefully be backported to older kernels.
As the kernel recommends, the resultant seeds are 256 bits and are
allocated using pool memory of type EfiACPIReclaimMemory, so that it
gets freed at the right moment in boot.
Let's make use of the new flag whenever we access the ESP or XBOOTLDR.
The resources we make use of in these partitions can't possibly use
symlinks (because UEFI knows no symlink concept), and they are untrusted
territory, hence under no circumstances we should be tricked into
following symlinks that shouldn't be there in the first place.
Of course, you might argue thta ESP/XBOOTLDR are VFAT and thus don#t
know symlinks. But the thing is, they don#t have to be. Firmware can
support other file systems too, and people can use efifs to gain access
to arbitrary Linux file systems from EFI. Hence, let's better be safe
than sorry.
I changed imports of util.h to initrd-util.h, or added an import of
initrd-util.h, to keep compilation working. It turns out that many files didn't
import util.h directly.
When viewing the patch, don't be confused by git rename detection logic:
a new .c file is added and two functions moved into it.
This moves the shim security arch override to the new
ReinstallProtocolInterface based interface. This also has the benefit to
reduce the time window in which we have this override active and also
actually removes it, which was not previously done.
The shim hooks themselves are also modernized too. The upcalls should
really not be neccessary if shim is happy with the provided binary.
Only the compat entry address is used now. This also now only returns
the compat entry address. If the image is native we do not need to try
calling into the entry address again as we would already have done so
from StartImage (and failed).
This is the proper way to start any EFI binary. The fact this even ever
worked was because the kernel does not have any PE relocations.
The only downside is that the embedded kernel image has to be signed and
trusted by the firmware under secure boot. A future commit will try to
deal with that.
This is really the parent image for the kernel that is to be run.
Renaming it as such prevents confusion with any image handles that are
about to be created.
This adds two more phases to the PCR boot phase logic: "sysinit" +
"final".
The "sysinit" one is placed between sysinit.target and basic.target.
It's good to have a milestone in this place, since this is after all
file systems/LUKS volumes are in place (which sooner or later should
result in measurements of their own) and before services are started
(where we should be able to rely on them to be complete).
This is particularly useful to make certain secrets available for
mounting secondary file systems, but making them unavailable later.
This breaks API in a way (as measurements during runtime will change),
but given that the pcrphase stuff wasn't realeased yet should be OK.
This fixes a regression introduced in e1636807 that removed setting this
value as it seemingly was not used by the kernel and would actively
break above 4G boots. But old kernels (4.18 in particular) will not boot
properly if it is not filled out by us.
The original issue was using the truncated value to then jump into the
kernel entry point, which we do not do anymore. So setting this value
again on newer kernels is fine.
The compiler may decide computations like these are not doing anything
and decide to optimize them away. This would defeat the whole fuzzing
exercise. This macro will force the compiler to materialize the value
no matter what. It should be less prone to accidents compared to using
log functions, which would either slow things down or still optimize the
value away (or simply move it into the if branch the log macros create).
The benefit over assert_se would be that no requirement is made on the
value itself. If we are fine getting a string of any size (including
zero), an assert_se would either create a noisy compiler warning about
conditions that would alawys be met or yet again optimize the whole
thing away.
Luca Boccassi
Jason A. Donenfeld
Lennart Poettering
Gerd Hoffmann
Yu Watanabe
Jan Janssen
Zbigniew Jędrzejewski-Szmek